Countervailing inequality effects of globalization and...
Transcript of Countervailing inequality effects of globalization and...
-
Working Paper SeriesDepartment of Economics
University of Verona
Countervailing inequality effects of globalization and renewableenergy generation in Argentina
Andrea Vaona
WP Number: 12 July 2013
ISSN: 2036-2919 (paper), 2036-4679 (online)
-
1
Countervailing inequality effects of globalization and renewable energy
generation in Argentina
Andrea Vaona
(Corresponding author)
Department of Economic Sciences
University of Verona
Viale dell’artigliere 19
37129 Verona
E-mail: [email protected]
Kiel Institute for the World Economy
Hindenburgufer 66
D-24105 Kiel
mailto:[email protected]
-
2
Countervailing inequality effects of globalization and renewable energy
generation in Argentina
Abstract
The present paper assesses the impacts of renewable energy
generation and globalization on income inequality in Argentina. We
make use of vector autoregression models. We find that
globalization and hydroelectric power increase inequality, while the
opposite holds true for other renewable energy sources. Several
robustness checks are considered. Policy implications are discussed
keeping into account the specific Argentinean context.
Keywords: Argentina, VAR, energy sources, inequality, globalization
JEL Codes: Q20, Q40, F60, F63, D63.
-
3
Introduction
Economists have traditionally been interested in the connection of energy and macroeconomic
magnitudes. A classical research topic is, for instance, the energy consumption-output nexus, which
recently took into consideration renewable energy sources too (see for example Bilgen et al., 2004; Apergis
and Payne, 2012, 2013)1. The present paper investigates a new topic: namely the effects of globalization
and renewable energy generation on income inequality. Interestingly, these phenomena have not been
considered together so far.
We chose a specific country for our analysis, namely Argentina. We did so for various reasons.
Among developing countries, Argentina has excellent information in the World Development Indicators
database with respect to the variables of interest. Furthermore this country has had a relevant weight in
the international economics debate since the 2002 peso crisis (Fanelli, 2002; Hayo and Neuenkirch, 2013).
In addition, in Argentina globalization first increased and then decreased. We hope that the variability of
our index of economic globalization will help us to better identify its effect on inequality. Furthermore, the
peso crisis was followed by a period of relevant economic growth (Herrera and Tavosnanska, 2011).
However, this revived growth did not fully translate into a more equitable society (Groisman, 2008).
Therefore, exploring further determinants of income inequality in the country under analysis is still
challenging. Finally, Argentina has a considerable unexploited potential with respect to renewable energy
generation, as it mainly relies on fossil fuels (Guzowski and Recalde, 2008; see also the descriptive statistics
below). So if we find that it can affect inequality, there will be one more important reason to encourage its
development.
The link between globalization on one side and inequality on the other has been widely explored.
Various literature surveys are available by now. Crinò (2009) focuses on studies concerning off-shoring and
wage inequality between skilled and unskilled workers, concluding that, in originating countries, the former
increases the latter one. Kurokawa (2012) reviews contributions on trade and wage inequality. This link is
the subject of heating debate as most of the literature considered skill biased technical change the main
1 Surveys are available in Lee (2005, 2006), Yoo (2006), Chontanawat et al. (2006) and Payne (2009, 2010a, 2010b) and
a meta-analysis in Bruns et al. (2013).
-
4
factor behind increasing inequality worldwide. Nonetheless, there are some contributions stressing the link
between trade and wage inequality. For instance, Feenstra and Hanson (1996) points to the fact that off-
shoring increases wage inequality in receiving countries too. This is because, by their standards, incoming
production activities are high skilled ones. This makes the wages of high- and low-skilled workers to
diverge. According to Dinopoulos and Segerstrom (1999), instead, trade impacts on technological progress.
It increases the reward for innovation, encouraging complex R&D activities both in developed and
developing countries. This increases the demand for high skilled workers worldwide, boosting their relative
wage with respect to low-skilled ones. Another example of model that revived the trade explanation for
increasing inequality is Kurokawa (2011). According to this work, the link between trade and inequality
passes through the increasing variety of goods demanded worldwide, which boosts the demand of high-
skilled workers compared to low-skilled ones. Also Chusseau et al. (2008) agree that both trade and
technological change have a role in explaining increasing wage inequality worldwide and they stress the
importance of models assessing the impact of trade on endogenous technological change.
The connection between different energy sources and inequality has received more attention in
policy oriented contributions than in quantitative economic studies. For instance, Sagar (2005) stresses that
traditional energy sources badly affects the well-being of poor households in developing countries. A fund
is proposed to help households to adapt to more modern energy carriers. Birol (2007) and Kaygusuz (2007)
highlight the political relevance of energy poverty in developing countries. Recently, Srivastava and Sokona
(2012) authoritatively brought again the issue to the attention of energy scholars and policy makers,
collecting various contributions dealing with energy poverty in different parts of the world. General themes
that emerge from these works are that solving the problem of energy access for poor households is a
complex task. Often, public discourses and inadequate monitoring tools prevent a sufficient public
awareness of the issues at stake. Developing countries should strive to find their own strategies, carefully
considering their own needs, by involving local populations in decision processes. Key factors are not only
building appropriate infrastructures, but also institutional frameworks that foster energy innovations, able
to overcome the rural/urban electricity divides that often exist in developing countries. In order to put in
-
5
place processes able to achieve all of these aims, government intervention is needed, not only in the form
of institution building but also through public investments, while preserving financial stability. The
importance of public-private partnerships should not be downplayed too.
Why energy poverty should be connected to income inequality though? This is for various reasons.
In the first place traditional energy carriers damage the health of people, which naturally become less
productive. Secondly, they are time consuming too. For instance, in developing countries, it is often the
case that women have to spend a considerable amount of their time in finding wood and in bringing it
home. Energy transition could free up time for other activities, both domestic and market ones. It could
also yield direct revenues and employment in new energy sectors, as well as the birth of new firms.
In our view, all the above reasons warrant a quantitative study gauging the effects of globalization
and the adoption of different energy sources on income inequality at the same time. Our method is the
estimation of vector autoregression (VAR) models, as detailed below.
The rest of this paper is structured as follows. First we describe our dataset and data sources. Next
we set out our results with respect to inequality, globalization and renewable energy sources. Then we
perform a series of robustness checks, such as including other possible explanatory variables and
considering other energy carriers. Finally, we conclude by contextualizing our results and the general
considerations contained in the present Introduction within the Argentinean energy system. Therefore, we
do not neglect that each country needs to follow its own path to energy transition, as also stressed by
Sokona et al. (2012). In the Appendix we provide further evidence that the variables under study are
stationary.
Dataset description
We start by focusing on three variables, the GINI index, which is our measure of inequality; the
percentage of electricity production from renewable sources excluding hydroelectric power (RENPERC); and
the KOF index of economic globalization after Dreher et al. (2008). The GINI index and RENPERC comes
from the World Economic Indicators of the World Bank. The KOF index of economic globalization takes into
-
6
account not only long distance flows of capital, goods and services, but also information and perceptions
arising through market exchanges. Detailed information on how it is computed is available at
http://globalization.kof.ethz.ch/.
Note that three observations are missing for the GINI index in 1988, 1989 and 1990. Therefore our
baseline estimates will be based on observations from 1991 to 2010. However, one of our robustness
checks will be to control that our results are not affected once using linear interpolation to fill the gap in
the sample, so extending it from 1986 to 2010.
In our robustness checks we consider further variables. We add to the model domestic credit
provided by the banking sector as percentage of GDP (PRIVATEBAN) and domestic credit to the private
sector as percentage of GDP (PRIVATE). We insert these variables in our model one at a time. This is
because Beck et al. (2007) found that financial development, measured in a similar way as above, can
reduce inequality in a cross-country panel data study.
Next we try to substitute RENPERC respectively with the percentage of electricity production from
hydroelectric sources (HYDROPERC), from nuclear sources (NUCLEARPERC) and from oil, gas and coal
sources (FOSSILPERC). This is to check whether our results hold also when considering other energy
carriers. PRIVATEBAN, PRIVATE, HYDROPERC, NUCLEARPERC and FOSSILPERC are all taken from the World
Development Indicators. Table 1 offers a glossary of all the acronyms used in the present work. Note that in
the rest of this paper when the acronym of a variable is preceded by an L, it means that it is considered in
natural logs.
Table 2 shows descriptive statistics of our variables. As it is possible to see there is a good variability
in the data, which is promising for our analysis. The next sections illustrate our results.
Results
We start by focusing on LGINI, LRENPERC, and LKOF. Our goal is, in the first place, to compute
impulse response functions to detect the effect of LRENPERC and LKOF on LGINI and the possible counter-
effects of the last variable on the first two. So we specify a VAR in levels.
http://globalization.kof.ethz.ch/
-
7
Table 1 - Glossary of the acronyms
Acronyms Explanation
GINI GINI index
GINIPOL Interpolated GINI index
RENPERC Electricity production from renewable sources excluding hydroelectric power (% of total) KOF KOF index of economic globalization
PRIVATE Natural log of domestic credit to private sector (% of GDP)
PRIVATEBAN Natural log of domestic credit provided by banking sector as % of GDP
HYDROPERC Electricity production from hydroelectric sources (% of total)
NUCLEARPERC Electricity production from nuclear sources (% of total)
FOSSILPERC Electricity production from oil, gas and coal sources (% of total)
Note: acronyms preceded by an L mark natural logarithms of the relevant variables.
Table 2 - Descriptive statistics
Acronyms Observations Mean Std. Dev. Maximum Minimum
GINI 22 48.34 3.18 54.72 42.79
GINIPOL 25 48.05 3.08 54.72 42.79
RENPERC 25 0.92 0.86 2.96 0.17
KOF 25 49.61 6.95 57.96 38.98
PRIVATE 25 18.16 6.23 39.72 10.49
PRIVATEBAN 25 35.89 13.22 80.06 22.48
HYDROPERC 25 34.35 5.95 43.56 25.01
NUCLEARPERC 25 9.64 2.70 14.46 5.72
FOSSILPERC 25 55.08 6.95 66.64 43.13
In order to decide how many lags to insert we rely on lag order selection criteria. The sequential
modified likelihood ratio (LR) statistic, the final prediction error, the Akaike, the Schwarz and the Hannan-
Quinn information criteria all point to one lag. We run a lag exclusion Wald test, which strongly rejects the
null that the first lag in levels of all variables should be dropped with p-values of order of 0.00.
Also block exogeneity Wald tests support our model given that for all the three equations the null
that regressors should be dropped is rejected with p-values smaller than 0.01. On the basis of residuals
inspection we insert three dummy variables for the years 1996, 2002 and 2006 respectively. In so doing we
-
8
try to deal with possible structural breaks in the data2. It is worth noting that 2002 was the year of the peso
crisis. 1996 and 2006 are the years in which electricity generation from renewable sources respectively
took off and peaked.
We focus on the residuals of our model by running serial correlation, normality and
heteroskedasticity tests. The Portmanteau test for autocorrelation with the presence of two lags of the
variables under study accepts the null of no serial correlation with a p-value of 0.33. The same does the
Lagrange Multiplier (LM) tests with one and two lags with p-values of respectively 0.12 and 0.65. Residuals
appear to be normally distributed as the Jarque-Bera test does not reject the null of normality with a p-
value of 0.75. The null of no heteroskedasticity is not rejected with a p-value of 0.53. So residuals of the
VAR(1) model are well behaved.
The roots of the system points to stability being all less than one in modulus: two of them are equal
to 0.96 and the other to 0.48. The stability of the system is confirmed by running Kwiatkowski, Phillips,
Schmidt and Shin (1992) tests as showed by Table 3.3
Table 3 - Kwiatkowski, Phillips, Schmidt and Shin (1992) tests for stationarity for variables in logs.
LGINI LKOF LRENPERC
Kwiatkowski-Phillips-Schmidt-Shin test statistic 0.231816 0.286258 0.111166 Asymptotic critical values*: 1% level 0.739000 0.739000 0.216000
5% level 0.463000 0.463000 0.146000 10% level 0.347000 0.347000 0.119000
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
Note: for all the variables we used a Newey-West automatic bandwidth and a Bartlett Kernel.
Exogenous terms include only the constant for LGINI and LKOF and the constant and a linear trend
for LRENPERC. Asymptotic critical values were taken from Table 1 of Kwiatkowski, Phillips, Schmidt
and Shin (1992).
We included a time trend in the testing model for LRENPERC. This choice was comforted by
econometric testing. The p-values of the time trend in the testing models for LGINI and LKOF were both
greater than 0.05, being respectively equal to 0.16 and 0.051. Furthermore, we run a number of panel unit
root tests on the residuals of the VAR(1) in levels with the three dummies above. The Levin, Lin and Chu
2 Looking for structural breaks in variable coefficients seems prohibitive given the sample size.
3 Recall that further evidence concerning the stationarity of variables is provided in the Appendix.
-
9
test, the Im, Pesaran and Shin test, the ADF - Fisher Chi-square test and the PP - Fisher Chi square test all
strongly reject the presence of unit roots in the residuals with p-values of the order of 0.004.
Further note that the R2 of the LGINI equation is equal to 0.85, the one of the LRENPERC equation
to 0.99 and the one of the LKOF equation to 0.92. Adjusted R2
respectively are 0.77, 0.99 and 0.88. The high
percentage of explained variance downplays the risk of omission of relevant variables. Finally, we tried to
insert a time trend in the VAR because we did so in testing for stationarity of LRENPERC. For all the three
equations its coefficients were never statistically different from zero. So we preferred the specification
without the time trend5.
On the basis of the above evidence we feel confident in estimating impulse response functions.
Results are displayed in Figures 1 and 26. As it is possible to see, LGINI does not appear to have either
statistically or economically significant effects on LRENPERC and LKOF, while an increase in economic
globalization and in the percentage of electricity produced by renewable sources respectively increases and
decreases inequality, as measured by the log of the GINI coefficient7. Considering only the LGINI equation,
the short run elasticity of LGINI with respect to LRENPERC is 0.01, while with respect to LKOF is 0.24. Long-
run elasticities respectively are 0.02 and 0.45. Given the above results we will not consider any more the
possible counter-effects of inequality on other variables in the rest of the paper.
4 We used the SIC criterion to select lag length and the the Newey - West automatic bandwidth using the Bartlett
kernel where appropriate. 5 T-statistics were equal to 0.33404, -0.99180 and -0.62839 for the LGINI, the LKOF and the LRENPERC respectively.
6 In all the impulse response functions displayed in this paper, we show only a limited amount of periods, the reason
being that we focus on time periods where the effect of impulses is statistically different from zero. 7 Note that our results would not change once using interpolation to fill the missing values in the GINI coefficients and
so extending our sample of reference. We preferred to be conservative and to rely on the smaller sample.
-
10
Figure 1 - Accumulated response functions to Cholesky one standard deviations
Note: dotted lines denote impulse responses plus and minus their standard errors.
-
11
Figure 2 - Accumulated response functions to Cholesky one standard deviations
Note: dotted lines denote impulse responses plus and minus their standard errors.
Robustness checks
Variables in levels
Our first robustness check is considering a different functional specification, namely we do not
apply the log transformation to our data. We estimate a VAR in GINI, RENPERC and KOF. On the basis of
residuals inspection we insert three dummies: for the years 2001, 2002 and 2006 respectively. Most of the
above results are confirmed. Roots are within the unit circle, being the greatest equal to 0.82 and the
smallest to 0.03. The stability of the system is again supported by the Kwiatkowski, Phillips, Schmidt and
Shin (1992) tests as documented in Table 4.
-
12
Table 4 - Kwiatkowski, Phillips, Schmidt and Shin (1992) tests for stationarity for variables in
levels GINI KOF RENPERC
Kwiatkowski-Phillips-Schmidt-Shin test statistic 0.273557 0.273558 0.128845 Asymptotic critical values*: 1% level 0.739000 0.739000 0.216000
5% level 0.463000 0.463000 0.146000 10% level 0.347000 0.347000 0.119000
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
Note: for LGINI and LKOF we used a Newey-West automatic bandwidth and a Bartlett Kernel. The
only exogenous term is a constant. For LRENPERC we used a Newey-West automatic bandwidth
and a Parzen Kernel. Exogenous terms include a constant and a linear trend. Asymptotic critical
values were taken from Table 1 of Kwiatkowski, Phillips, Schmidt and Shin (1992).
The sequential modified LR statistic points to one lag length, but all the other lag length criteria -
the final prediction error, the Akaike, the Schwarz and the Hannan-Quinn information criteria - point to
two. We run a lag exclusion Wald test, which strongly rejects the null that both the first and second lags of
all variables should be dropped with p-values of order of 0.00. Therefore we prefer a VAR(2) model in this
case.
The Portmanteau test for autocorrelation with the presence of three lags of the variables under
study accepts the null of no serial correlation with a p-value of 0.09. The same does the LM tests with one,
two and three lags with p-values of respectively 0.10, 0.79 and 0.33. Residuals appear to be normally
distributed as the Jarque-Bera test does not reject the null of normality with a p-value of 0.39. The null of
no heteroskedasticity is not rejected with a p-value of 0.19. So residuals of the VAR(2) model are well
behaved.
Once again panel unit root tests on the residuals of the VAR(2) in levels support the model. The
Levin, Lin and Chu test, the Im, Pesaran and Shin test, the ADF - Fisher Chi-square test and the PP - Fisher
Chi square test all strongly reject the presence of unit root in the residuals with p-values of the order of
0.008.
We try to use a different definition of impulse response functions, now considering responses to
non-factorized one unit innovations as displayed in Figure 3. Once again an increase in the percentage of
8 Again we used the SIC criterion to select lag length and the Newey - West automatic bandwidth using the Bartlett
kernel where appropriate.
-
13
electricity produced by renewable energy sources significantly decreases inequality and the opposite
happens for globalization.
Figure 3 - Responses to non-factorized one unit innovations
Note: dotted lines denote impulse responses plus and minus their standard errors.
Adding indicators of financial development
Our next robustness check consists in adding indicators of financial development, customarily used
in the finance-growth literature and that Beck et al. (2007) found to reduce inequality. We start with
LPRIVATEBAN. We do so by sticking to our baseline estimates. All tests would support the model, but the
block exogeneity Wald test which returns a p-value of 0.24 for the LPRIVATEBAN equation. Therefore we
-
14
specify the model in a different way. We drop the bespoken equation and we insert the first lag of
LPRIVATEBAN as an exogenous regressor in a VAR(1) of LGINI, LRENPERC and LKOF.
Figure 4 - Accumulated Response to Non-factorized One S.D. Innovations ± 2 S.E.
Note: dotted lines denote impulse responses plus and minus their standard errors.
Block exogeneity tests now report p-values smaller than 0.05 for all the equations. Roots are within
the unit circle, being the greatest two equal to 0.98 and the smallest one to 0.05. All length criteria prefer a
value of 1. The null that this lag is equal to zero is again rejected with a p-value of 0.00. The presence of
serial correlation is rejected by the Portmanteau test for autocorrelation with the presence of two and
-.08
-.04
.00
.04
.08
.12
.16
1 2 3 4 5 6 7 8 9 10
Accumulated Response of LGINI to LRENPERC
-.08
-.04
.00
.04
.08
.12
.16
1 2 3 4 5 6 7 8 9 10
Accumulated Response of LGINI to LKOF
-
15
three lags, being p-values respectively equal to 0.40 and 0.67. LM tests with one, two and three lags report
p-values of respectively 0.12, 0.81 and 0.95. Normality is not rejected as the Jarque-Bera estimator has a p-
value of 0.98. The heteroskedasticity test returns a p-value of 0.27.
For robustness sake we choose for the present case, yet a different definition of impulse response
function, namely the accumulated response to non-factorized one standard deviation innovations. Figure 4
shows that increasing LKOF increases LGINI, while increasing LRENPERC decreases LGINI. The effect of
LPRIVATEBAN does not turn out to be statistically significant, as all its coefficients in the three VAR
equations do not display p-values smaller than 0.05.
Suppose to consider instead variables in levels, without the log transformation. We estimate a VAR
involving the variables GINIPOL, KOF, RENPERC and PRIVATE. Once again a block exogeneity test leads to
drop the equation for PRIVATE, whose first lag is then included among exogenous variables. All lag length
criteria point to 1 lag as the most suitable specification. Both the Portmanteau and the LM tests for
autocorrelation find hardly any evidence of serial correlation. The former returns a p-value of 0.873 with 2
lags and the latter of 0.86 with one lag. Therefore we stick to a VAR(1). The Jarque-Bera test returns a p-
value of 0.07 supporting normality and the residual heteroskedasticity term a p-value of 0.59, not rejecting
the null of absence of heteroskedasticity.
Roots are within the unit circle as two of them are in modulus equal to 0.87 and the other one to
0.46. Variables appear to be stationary according to the Kwiatkowski, Phillips, and Schmidt and Shin (1992)
test as showed in Table 5. Once again the panel unit root tests mentioned above do not find any evidence
of unit roots, all returning p-values of 0.00 once applied to the residuals of the three equations of the VAR
under study.
PRIVATE is not statistically significant in any of the VAR equations, so it can be dropped. Once
moving to inspect responses to Cholesky one standard deviations innovations we find once again that
inequality is positively connected to economic globalization and negatively to renewable energy electricity
-
16
generation as percentage of total generation (Figure 5). We next move on to consider whether it is possible
to find similar results to those above for other energy sources.
Table 5 - Kwiatkowski, Phillips, Schmidt and Shin (1992) tests for stationarity for variables in
levels GINIPOL KOF RENPERC PRIVATE
Kwiatkowski-Phillips-Schmidt-Shin test statistic 0.212311 0.196445 0.422081 0.326845 Asymptotic critical values*: 1% level 0.739000 0.739000 0.739000 0.739000
5% level 0.463000 0.463000 0.463000 0.463000 10% level 0.347000 0.347000 0.347000 0.347000
*Kwiatkowski-Phillips-Schmidt-Shin (1992, Table 1)
Note: for all variables we used a Newey-West automatic bandwidth and a Parzen Kernel. The only
exogenous term is a constant. Asymptotic critical values were taken from Table 1 of Kwiatkowski,
Phillips, Schmidt and Shin (1992).
Figure 5 - Response to Cholesky one standard deviation innovation
Note: dotted lines denote impulse responses plus and minus their standard errors.
-0.4
0.0
0.4
0.8
1.2
1.6
1 2 3 4 5 6 7 8 9 10
Response of GINIPOL to KOF
-0.4
0.0
0.4
0.8
1.2
1.6
1 2 3 4 5 6 7 8 9 10
Response of GINIPOL to RENPERC
-
17
Other energy sources
First we start with hydroelectric power, then we move on to nuclear energy and finally we consider
fossil sources. We first start to test for the stationarity of HYDROPERC, FOSSILPERC and NUCLEARPERC.
Kwiatkowski, Phillips, Schmidt and Shin (1992) tests accept the null at least at the 5% level for all the
variables under analysis (Table 6).
Table 6 - Kwiatkowski, Phillips, Schmidt and Shin (1992) tests for stationarity for variables in
levels HYDROPERC FOSSILPERC NUCLEARPERC
Kwiatkowski-Phillips-Schmidt-Shin test statistic 0.252860 0.116454 0.427933 Asymptotic critical values*: 1% level 0.739000 0.216000 0.739000
5% level 0.463000 0.146000 0.463000 10% level 0.347000 0.119000 0.347000
Note: for all variables we used a Newey-West automatic bandwidth and a Parzen Kernel. The only
exogenous term is a constant, with the exception of FOSSILPERC where a trend was inserted on the
basis of its statistical significance in the model underlying the stationarity test. Asymptotic critical
values were taken from Table 1 of Kwiatkowski, Phillips, Schmidt and Shin (1992).
Hydroelectric generation
We estimate a VAR with GINI, HYDROPERC and KOF. We start with a VAR(1) model, which all tests
on residuals, stationarity and lag length would support. However, block exclusion Wald tests return a p-
value of 0.00 for the GINI equation, one of 0.81 for the HYDROPERC equation and one of 0.07 for the KOF
equation. Therefore we re-specify the model considering the first lag of HYDROPERC as an exogenous
variable. The block exclusion Wald test of the KOF equation still returns a p-value of 0.98.
Therefore we regressed GINI on a constant its first lag, and the first lags of KOF and HYDROPERC.
We used White heteroskedasticity consistent standard errors. The Breusch-Godfrey serial correlation LM
test with 2 and 14 degrees of freedom reports a p-value of 0.29. Therefore we do not insert further lags of
variables in the model. The Jarque-Bera normality test for the residuals displays a p-value of 0.85. The
RESET test does not find evidence of omitted variables - as inserting the square of the fitted values of the
model, it returns a p-value of 0.16. The CUSUM and CUSUM of squares considering all coefficients do not
detect any presence of structural breaks as showed by Figure 6.
In brief our estimated model is
GINI = 8.97+ 0.19*KOF(-1) + 0.12*HYDROPERC(-1) + 0.53*GINI(-1) + ut (1)
GINI = (0.24) (0.04)*KOF(-1) (0.04)*HYDROPERC(-1) (0.02)*GINI(-
-
18
where p-values are reported in parentheses and ut is a stochastic error. This model has a R2 of 0.83 and an
adjusted-R2 of 0.80. The above coefficient estimates imply short-run elasticities at mean values equal to
0.20 and 0.08 for KOF and HYDROPERC respectively. Respective long-run elasticities are 0.42 and 0.17.
Figure 6 - Structural breaks test for equation 1.
-12
-8
-4
0
4
8
12
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM 5% Significance
-0.4
0.0
0.4
0.8
1.2
1.6
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM of Squares 5% Significance
-
19
Nuclear electricity generation
We estimate a VAR with GINI, NUCLEARPERC and KOF. All tests would support a VAR(1) model.
However Wald block exogeneity tests lead to specify a single equation as happened for HYDROPERC.
GINI was regressed on a constant, its first lag, and the first lags of KOF and HYDROPERC, by using
White heteroskedasticity consistent standard errors. No serial correlation was detected by the Breusch-
Godfrey test with 2 and 14 degrees of freedom, which reported a p-value of 0.40. No further lags were
inserted therefore in the model. Normality was not rejected by the Jarque-Bera test displaying a p-value of
0.54. The square of the fitted values, once inserted in the model, has a p-value of 0.16, not letting to
suppose the omission of any relevant variable. No structural break was detected by the CUSUM and the
CUSUM of squares tests as showed by Figure 7.
Our estimated model, with p-value in parentheses, is
GINI = 10.61+ 0.19*KOF(-1) + 0.06*NUCLEARPERC(-1) + 0.56*GINI(-1) + et (2)
GINI = (0.28) (0.04)*KOF(-1) (0.56) *HYDROPERC(-1) (0.04)*GINI(-
where et is a stochastic error. The R2 and the adjusted-R
2 respectively are 0.78 and 0.74. NUCLEARPERC
does not appear to have any significant impact on inequality.
Fossil electricity generation
We estimate a VAR with GINI, FOSSILPERC and KOF. We include a trend because we did so in the
stationarity test for FOSSILPERC. As with HYDROPERC and NUCLEARPERC block exogeneity tests would lead
to specify a single equation for GINI. We proceeded as before, performing a regression of GINI on a
constant, the first lags of GINI, KOF and FOSSILPERC with White heteroskedasticity consistent standard
errors. No serial correlation was detected as the Breusch-Godfrey test with 2 and 14 degrees of freedom
reported a p-value of 0.28. Figure 8 shows the absence of structural breaks. The null of normality was
accepted with a p-value of 0.82. The square of the fitted values, once inserted in the model, has a p-value
of 0.22. FOSSILPERC had a statistically insignificant coefficient of -0.09 with a p-value of 0.09. The other
coefficients and p-values, as well as the R2 and the adjusted-R
2 were of the same order of magnitude as in
-
20
the two models above. Including a trend in the model would produce a insignificant coefficient of -0.03
with a p-value of 0.47.9
Figure 7 - Structural breaks test for equation (2)
9 As a final robustness check suppose to insert LHYDROPERC, LNUCLEARPERC and LFOSSILPERC in our baseline VAR.
They would never have significant coefficient in any model. More details are available from the author upon request.
-12
-8
-4
0
4
8
12
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM 5% Significance
-0.4
0.0
0.4
0.8
1.2
1.6
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM of Squares 5% Significance
-
21
Figure 8 - Structural breaks test for the model including FOSSILPERC
-12
-8
-4
0
4
8
12
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM 5% Significance
-0.4
0.0
0.4
0.8
1.2
1.6
95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10
CUSUM of Squares 5% Significance
-
22
Conclusions and implications for energy policy in Argentina
In brief, our results are that in Argentina globalization and hydroelectric power increase inequality
while other renewable energy sources decrease it. Further energy sources do not affect inequality. The
effect of globalization is larger than those of energy sources. This means that there should be a
considerable increase in renewable energy generation (excluding hydroelectric power) to offset the
consequences of globalization on income inequality.
Regarding the result on globalization, we cannot offer a test for the various mechanisms the
literature proposed to explain the effect of globalization on inequality, because our index is a catch all
variable. However, we can confirm that in general globalization increases inequality, which is important
because this means that in our case the effect microeconomic mechanisms also show up once considering
econometric tests on macroeconomic magnitudes.10
Concerning hydroelectric power, one needs to remember that hydroelectric power is generated by
large dams in Argentina (Guzowski and Recalde, 2010, p. 5815). This kind of infrastructure has often been
very controversial. They have not only benefits, but also high environmental and social costs for local
communities. Flood control, water supply, low-cost energy and increased opportunities for recreation can
be provided. Hydro power also reduces CO2 emissions. However, large dams change a terrestrial ecosystem
into an aquatic one, often displacing people from the area to be inundated (Koch, 2002; Frey and Linke,
2002; Magnani, 2012, 9-13). Under extreme circumstances they can be a challenge to the security of local
communities, as the Frìas disaster in 1970 reminds us (see among others Singh, 1996, 85-86). What is more,
this kind of facilities mainly serves urban centers, so they often offer little help in overcoming rural/urban
divides (Lerer and Scudder, 1999). Therefore, it is not so surprising that hydropower generation by large
dams can increase inequality once taking into account the considerations above.
Regarding the effect of other renewable energy sources on inequality one needs to carefully
consider the Argentinean context. One first important thing to keep into account is that in Argentina,
according to the US Energy Information Administration, other renewable energy sources for electricity
10
For a similar point in a different field see Brown et al. (2009).
-
23
generation than hydroelectric power are wind (3%) and mostly biomass and waste (97%). As already
noticed both account for a very small share of overall electricity generation. What is more, though this
share tended to increase between 1994 and 2006, this trend was then reversed - especially for wind power,
whose share dropped back to its early nineties level.
Many studies concerned with electricity issues in Argentina focus on market reforms. Before the
2002 peso crisis, the country experienced deep deregulation and privatization processes, which were later
reversed. A price cap was even introduced. The relevant literature is divided, as some scholars claim that
deregulation and privatization were appropriate policies, while others highlight the pitfall of these policies
(see for instance Dutt et al., 1997; Pollit, 2008; Nagayama and Kashiwagi, 2007; Haselip, 2005; Haselip et
al., 2005, Haselip and Potter, 2010; Guzowski and Recalde, 2010).
However, different energy sources are an important issue in itself though. This is because a country
should not concentrate only on a single kind of renewable energy. On the contrary, diversification among
different renewable energy sources can also help overcoming intermittency problems. Policies to foster
renewable energy sources have been hampered so far by a large endowment of natural gas in Argentina
(Guzowski and Recalde, 2008). At present policy intervention is based on quota systems and subsidies,
however, according to some experts, these tools are less promising than feed-in tariffs, that were
successfully adopted in Germany, Spain and the US, for instance (Guzowski and Recalde, 2010).
Wind energy has a great potential in Argentina, but installed capacity is tiny and mainly in small
cooperative projects with the aim to supply rural local communities. Yet, this offers an example of the
mechanisms highlighted in the introduction whereby renewable energy generation can reduce inequality.
This notwithstanding current regulation and distorted prices put wind power at disadvantage and prevent
Argentina to fully exploit the benefits that other countries already enjoy (Guzowski and Recalde, 2008;
2010; Recalde, 2010; Kaygusuz, 2004).
As successful countries in promoting renewable energy sources - such as Germany and Denmark -
different forms of government intervention are needed in Argentina too (Guzowski and Recalde, 2008;
-
24
2010). They should be directed at building a clear regulatory framework; promoting long-term contracts
and concessions to abate market risk; weaving national-international nets; supporting educational and
information initiatives; and, finally, favoring the access to the credit market by providing some form of
guarantees to entrepreneurs (Guzowski and Recalde, 2008; Larson and Kartha, 2000; Lawand and Ayoub,
1996). Many policy interventions of this kind were taken in Chile, for instance (Guzowski and Recalde,
2010). It is important to recall, though, that each single policy tool should be assessed within the overall
policy framework in place in a given country or region and together with its interaction with technological
progress and social factors (Fischer and Preonas, 2010; Escalante et al., 2013).
Finally, the advantages of renewable energy sources discussed in the introduction exist for biomass
energy too. However, for this kind of energy, there exists a peculiar trade-off, namely the competition for
land with the production of food. Furthermore, the increase in biomass energy generation is often achieved
by ousting tenant farmers and appropriating common lands (Larson and Kartha, 2000). In the past, there
was little attention to these issues. However, there exists ways to soften the bespoken trade-off thanks to
technological advances in agricultural activities and with a better coordination of international, national
and especially local institutions. The last ones and local population at large should have the chance to
express and defend local needs. Methods to enable this have been studied in the literature (Muguerza et
al., 1990; García-López and Arizpe, 2010). This is important for various reasons. It increases the chances of
success for the various projects, which can better adapt to local complexities. Furthermore, it can also
prevent deforestation and preserve local biodiversity (Larson and Williams, 1995; Larson and Kartha, 2000).
The relevance of these issues should not be downplayed, because what is at stake is the possibility to
assure not only the carbon-neutrality of biomass energy, but also the chance to make it carbon negative
(Mathews and Goldsztein, 2009; Manrique et al, 2011). Moreover forests can be a source of energy in
themselves, once using wood pellets and sawdust produced by the wood industry as energy generation
-
25
inputs (Uasuf and Becker, 2011; Bandyopadhyay, 2011)11
. Hopefully, international policy makers seem to be
aware of the issues at stake, as testified by the PROBIOMASA FAO project.12
References
Andrews D. (1993) Exactly median unbiased estimation of first order autoregressive/unit root
models, Econometrica 61, 139-165.
Andrews, D. and H.Y. Chen. 1994. Approximately Median Unbiased Estimation of Autoregressive
Models. Journal of Business and Economic Statistics 12: 187-204
Apergis, N. and J. E. Payne. 2012. "A Global Perspective on the Renewable Energy Consumption-
Growth Nexus", Energy Sources, Part B: Economics, Planning, and Policy, 7, 314-322.
Apergis, N. and J. E. Payne. 2013. "Another Look at the Electricity Consumption-Growth Nexus in
South America", Energy Sources, Part B: Economics, Planning, and Policy, 8, 171-178
Bandyopadhyay Sushenjit, Priya Shyamsundar, Alessandro Baccini, Forests, biomass use and
poverty in Malawi, Ecological Economics, Volume 70, Issue 12, 15 October 2011, Pages 2461-2471, ISSN
0921-8009, http://dx.doi.org/10.1016/j.ecolecon.2011.08.003.
Beck Thorsten & Asli Demirgüç-Kunt & Ross Levine, 2007. "Finance, inequality and the poor,"
Journal of Economic Growth, Springer, vol. 12(1), pages 27-49, March
Bilgen, S., Kaygusuz, K. and A. Sari. 2004 "Renewable Energy for a Clean and Sustainable Future",
Energy Sources, 26, 12, 1119-1129.
Birol Fatih, 2007."Energy Economics: A Place for Energy Poverty in the Agenda?" The Energy
Journal, International Association for Energy Economics, vol. 0(Number 3), pages 1-6.
11
A series of specific proposals to improve the sustainability of the biomass industry in Argentina was advanced in
Shafik et al. (2006). 12
http://ccap.org/assets/Argentina_Renewable_Energy_Probiomassa_May_2013_NAMA_Executive_Summary.pdf
http://ccap.org/assets/Argentina_Renewable_Energy_Probiomassa_May_2013_NAMA_Executive_Summary.pdf
-
26
Brown, J. R., Steven M. Fazzari and Bruce C. Petersen, 2009. "Financing Innovation and Growth:
Cash Flow, External Equity, and the 1990s R&D Boom." Journal of Finance 64: 151-185.
Bruns, Stephan B., Gross, Christian and Stern, David I., Is There Really Granger Causality between
Energy Use and Output? (March 8, 2013). Crawford School Research Paper No. 13-07. Available at SSRN:
http://ssrn.com/abstract=2232455 or http://dx.doi.org/10.2139/ssrn.2232455
Chontanawat, J., Hunt, L.C., Pierse, R., 2006. Causality between energy consumption and GDP:
evidence from30 OECD and 78 non-OECD countries. Paper No. SEEDS 113. Surrey Energy Economics
Discussion Paper Series. University of Surrey.
Chusseau, N., Dumont, M. and Hellier, J. (2008), EXPLAINING RISING INEQUALITY: SKILL-BIASED
TECHNICAL CHANGE AND NORTHにSOUTH TRADE. Journal of Economic Surveys, 22: 409に457. doi:
10.1111/j.1467-6419.2007.00537.x
Crinò, Rosario (2009), Offshoring, Multinationals and Labour Market: A Review of the Empirical
Literature, Journal of Economic Surveys, 2009, 23(2), pp. 197-249
Dinopoulos, Elias & Segerstrom, Paul, 1999. "The dynamic effects of contingent tariffs," Journal of
International Economics, Elsevier, vol. 47(1), pages 191-222, February
Dreher, Axel, Noel Gaston and Pim Martens (2008), Measuring Globalisation に Gauging its
Consequences (New York: Springer).
Dutt Gautam S., Fernando G. Nicchi, Mario Brugnoni, Power sector reforms in Argentina: an update,
Energy for Sustainable Development, Volume 3, Issue 6, March 1997, Pages 36-54, ISSN 0973-0826,
http://dx.doi.org/10.1016/S0973-0826(08)60219-7.
Escalante, K.N. & Belmonte, S. & Gea, M.D., 2013. "Determining factors in process of socio-
technical adequacy of renewable energy in Andean Communities of Salta, Argentina," Renewable and
Sustainable Energy Reviews, Elsevier, vol. 22(C), pages 275-288.
http://ideas.repec.org/s/bla/jfinan.html
-
27
Fanelli, José María. 2002, "Growth, instability and the convertibility crisis in Argentina". CEPAL
Review, 77, 25-43.
Feenstra, Robert C & Hanson, Gordon H, 1996. "Globalization, Outsourcing, and Wage Inequality,"
American Economic Review, American Economic Association, vol. 86(2), pages 240-45, May.
Fischer, C., Preonas, L., 2010. Combining policies for renewable energy. Discussion Paper 10-19.
Resources for the future, Washington DC.
Frey, Gary W. and Deborah M. Linke (2002), Hydropower as a renewable and sustainable energy
resource meeting global energy challenges in a reasonable way, Energy Policy 30, 1261に1265.
García-López Gustavo A., Nancy Arizpe, Participatory processes in the soy conflicts in Paraguay and
Argentina, Ecological Economics, Volume 70, Issue 2, 15 December 2010, Pages 196-206, ISSN 0921-8009,
http://dx.doi.org/10.1016/j.ecolecon.2010.06.013.
Groisman, Fernando, 2008. "Distributive effects during the expansionary phase in Argentina (2002-
2007)". CEPAL Review, 96, 203-222.
Guzowski C., M. Recalde, Latin American electricity markets and renewable energy sources: The
Argentinean and Chilean cases, International Journal of Hydrogen Energy, Volume 35, Issue 11, June 2010,
Pages 5813-5817, ISSN 0360-3199, http://dx.doi.org/10.1016/j.ijhydene.2010.02.094.
Guzowski, C., M. Recalde, Renewable energy in Argentina: Energy policy analysis and perspectives,
International Journal of Hydrogen Energy, Volume 33, Issue 13, July 2008, Pages 3592-3595, ISSN 0360-
3199, http://dx.doi.org/10.1016/j.ijhydene.2007.11.032.
Haselip, J. A. (2005). Renegotiating Electricity Contracts after an Economic Crisis and Currency
Devaluation: The Case of Argentina. The Electricity Journal, 18(3), 78-88 doi: 10.1016/j.tej.2005.03.003
Haselip, J. A., Dyner, I., & Cherni, J. (2005). Electricity market reform in Argentina: assessing the
impact for the poor in Buenos Aires. Utilities Policy, 13(1), 1-14 doi: 10.1016/j.jup.2004.03.001
-
28
Haselip, James & Potter, Clive, 2010, "Post-neoliberal electricity market 're-reforms' in Argentina:
Diverging from market prescriptions?," Energy Policy, Elsevier, vol. 38(2), pages 1168-1176, February.
Hayo Bernd and Matthias Neuenkirch (2013), Does the currency board matter? US news and
Argentine financial market reaction, Applied Economics, 45, 4034-4040.
Herrera, Germán and Andrés Tavosnanska, 2011. "Argentine industry in the early twenty-first
century (2003-2008)", CEPAL Review, 104, 99-117.
Kaygusuz Kamil (2004), Wind Energy: Progress and Potential, Energy Sources, 26, 95-105.
Kaygusuz, K., 2007. Energy for sustainable development: key issues and challenges. Energy Sources,
Part B, 2, 73に83.
Koch, Frans H. (2002). Hydropower ね the politics of water and energy: Introduction and overview,
Energy Policy 30, 1207に1213.
Kurokawa Yoshinori, 2011. "Variety-skill complementarity: a simple resolution of the trade-wage
inequality anomaly," Economic Theory, Springer, vol. 46(2), pages 297-325, February.
Kurokawa, Y. (2012), A Survey of Trade and Wage Inequality: Anomalies, Resolutions and New
Trends. Journal of Economic Surveys. doi: 10.1111/joes.12007.
Kwiatkowski, D., P. C. B. Phillips, P. Schmidt, and Y. Shin (1992): Testing the Null Hypothesis of
Stationarity against the Alternative of a Unit Root. Journal of Econometrics 54, 159に178.
Larson, E.D. and R.H. Williams. (1995). Biomass Plantation Energy Systems and Sustainable
Development. In Energy as an Instrument for Socio-Economic Development, J. Goldemberg and T.B.
Johansson, eds., 91-106, New York, United Nations Development Program.
Larson, Eric D. and Sivan Kartha, Expanding roles for modernized biomass energy, Energy for
Sustainable Development, Volume 4, Issue 3, October 2000, Pages 15-25, ISSN 0973-0826,
http://dx.doi.org/10.1016/S0973-0826(08)60250-1.
-
29
Lawand T.A., J. Ayoub, Renewable energy activities in rural Argentina ね Educational aspects,
Renewable Energy, Volume 9, Issues 1に4, SeptemberにDecember 1996, Pages 1194-1198, ISSN 0960-1481,
http://dx.doi.org/10.1016/0960-1481(96)88491-6.
Lee, C.-C., 2005. Energy consumption and GDP in developing countries: a cointegrated panel
analysis. Energy Economics 27 (3), 415に427.
Lee, C.-C., 2006. The causality relationship between energy consumption and GDP in G-11 countries
revisited. Energy Policy 34, 1086に1093.
Lerer Leonard B, Thayer Scudder, Health impacts of large dams, Environmental Impact Assessment
Review, Volume 19, Issue 2, March 1999, Pages 113-123, ISSN 0195-9255,
http://dx.doi.org/10.1016/S0195-9255(98)00041-9.
Magnani Natalia 2012, The Green Energy Transition. Sustainable Development or Ecological
Modernization? "Sociologica", 2, 1-25, http://www.sociologica.mulino.it/doi/10.2383/38270.
Manrique Silvina, Judith Franco, Virgilio Núñez, Lucas Seghezzo, Potential of native forests for the
mitigation of greenhouse gases in Salta, Argentina, Biomass and Bioenergy, Volume 35, Issue 5, May 2011,
Pages 2184-2193, ISSN 0961-9534, http://dx.doi.org/10.1016/j.biombioe.2011.02.029.
Mathews John A., Hugo Goldsztein, Capturing latecomer advantages in the adoption of biofuels:
The case of Argentina, Energy Policy, Volume 37, Issue 1, January 2009, Pages 326-337, ISSN 0301-4215,
http://dx.doi.org/10.1016/j.enpol.2008.07.022.
Muguerza Daniel, Daniel Bouille, Erik Barney, A method for the appraisal of alternative electricity
supply options applied to the rural areas of Misiones Province, Argentina, World Development, Volume 18,
Issue 4, April 1990, Pages 591-604, ISSN 0305-750X, http://dx.doi.org/10.1016/0305-750X(90)90074-8.
-
30
Nagayama Hiroaki, Takeshi Kashiwagi, Evaluating electricity sector reforms in Argentina: lessons for
developing countries?, Journal of Cleaner Production, Volume 15, Issue 2, 2007, Pages 115-130, ISSN 0959-
6526, http://dx.doi.org/10.1016/j.jclepro.2005.11.056.
Orcutt G. H. and Winokur H.S. (1969) First order autoregressions: inference, estimation and
prediction, Econometrica 37, 1-14.
Payne, J. E., 2009. On the dynamics of energy consumption and output in the US, Applied Energy,
Volume 86, Issue 4, Pages 575-577.
Payne, J.E., 2010a. Survey of the international evidence on the causal relationship between energy
consumption and growth. Journal of Economic Studies 37, 53に95.
Payne, J.E., 2010b. A survey of the electricity consumption-growth literature. Applied Energy 87,
723に731.
Pollitt Michael, Electricity reform in Argentina: Lessons for developing countries, Energy Economics,
Volume 30, Issue 4, July 2008, Pages 1536-1567, ISSN 0140-9883,
http://dx.doi.org/10.1016/j.eneco.2007.12.012.
Quenouille M. H. (1956) Notes on bias estimation, Biometrika 43, 353-360.
Recalde, M., Wind power in Argentina: Policy instruments and economic feasibility, International
Journal of Hydrogen Energy, Volume 35, Issue 11, June 2010, Pages 5908-5913, ISSN 0360-3199,
http://dx.doi.org/10.1016/j.ijhydene.2009.12.114.
Sagar Ambuj D., Alleviating energy poverty for the world's poor, Energy Policy, Volume 33, Issue 11,
July 2005, Pages 1367-1372, ISSN 0301-4215, http://dx.doi.org/10.1016/j.enpol.2004.01.001.
Shafik Asal, Rémi Marcus, Jorge A. Hilbert, Opportunities for and obstacles to sustainable biodiesel
production in Argentina, Energy for Sustainable Development, Volume 10, Issue 2, June 2006, Pages 48-58,
ISSN 0973-0826, http://dx.doi.org/10.1016/S0973-0826(08)60531-1.
-
31
Singh, V.P. (1996), Dam Breach Modeling Technology, Kluwer Academic Publishers, Dordrecht, The
Netherlands.
Sokona Youba, Yacob Mulugetta, Haruna Gujba, Widening energy access in Africa: Towards energy
transition, Energy Policy, Volume 47, Supplement 1, June 2012, Pages 3-10, ISSN 0301-4215,
http://dx.doi.org/10.1016/j.enpol.2012.03.040.
Srivastava Leena and Youba Sokona, eds. (2012), Universal access to energy: Getting the framework
right, Energy Policy 47, Supplement 1, 1-94.
Uasuf Augusto, Gero Becker, Wood pellets production costs and energy consumption under
different framework conditions in Northeast Argentina, Biomass and Bioenergy, Volume 35, Issue 3, March
2011, Pages 1357-1366, ISSN 0961-9534, http://dx.doi.org/10.1016/j.biombioe.2010.12.029.
Yoo, S.-H., 2006. The causal relationship between electricity consumption and economic growth in
the ASEAN countries. Energy Policy 34 (18), 3573に3582.
Appendix
The present appendix shows further results confirming the stationarity of the series analysed in this
study. OLS estimates of autoregressive parameters in autoregressive models are well known to be biased
(see for example Orcutt and Winokur, 1969 or Quenouille, 1956). We show here that our results are robust
to adopting the exactly median unbiased (EMU) estimator proposed by Andrews (1993) for autoregressive
models of order 1. We do not need to resort to the Andrews and Chen (1994) estimator for autoregressive
processes of higher order, because both the Schwarz and the Hannan-Quinn information criteria point to
one lag as the most suitable specification for all the variables we considered13
.
13
Further details are available from the author upon request.
http://dx.doi.org/10.1016/j.biombioe.2010.12.029